Sony Music Has Now Asked Platforms to Remove More Than 260,000 AI Deepfake Tracks - Nearly Double Its March Count

Sony Music Entertainment has asked streaming and social platforms to remove more than 260,000 AI-generated "deepfake" tracks that imitate the voices and likenesses of the artists it represents, according to figures the company supplied to the Financial Times and confirmed to Billboard. The tally, current through the end of September, has almost doubled in six months.

The trajectory is the story. Sony disclosed roughly 135,000 takedown requests at the end of March 2026. Between April and September, the label says it filed more than 20,000 requests per month, adding at least 125,000 new cases in two quarters. The longer arc is steeper still: about 10,000 requests by November 2023, around 75,000 by March 2025, then the near-doubling in the most recent six months.

The count covers unauthorized impersonations, not AI music in general: songs that mimic a specific artist's voice, name or other identifiable traits without permission. Sony has named Harry Styles, Adele, Britney Spears, Queen and Michael Jackson among the artists whose voices have been copied. Importantly, the figures count requests filed, not confirmed removals - the FT report did not identify which platforms received the notices or how quickly they acted.

"It's happening at scale, and harms artists, misleads fans and undermines the ethical use of AI," a Sony Music spokesperson told Billboard. "Every fraudulent stream takes money away from the artists and songwriters whose music is being listened to. We will keep enforcing our artists' rights and expect platforms to act quickly and to prevent the same content from reappearing, but it's becoming an uphill struggle."

The supply side explains the struggle. French streaming service Deezer has reported that more than 90,000 fully AI-generated tracks are uploaded to its platform every day; only a fraction are impersonations of real artists, but the volume shows why manual review cannot keep pace. Sony's core complaint is that removal is a treadmill - the same or near-identical material can resurface within hours, forcing rights teams to re-flag content they have already taken down.

The legal backdrop remains unsettled in the United States. There is still no federal law that protects a person's name, image, likeness or voice from unauthorized AI deepfakes; protection depends on a patchwork of state statutes. The NO FAKES Act, which would create a federal authorization right for voice and likeness, reached the Senate floor last week with near-unanimous support but was blocked from unanimous-consent passage by Senator Ted Cruz, who cited concerns about the bill's treatment of satire and political commentary.

For now, the burden stays where it has been: with rights holders filing notice after notice, and with platforms under pressure to move detection upstream before upload rather than react track by track. Sony's numbers suggest the voluntary notice-and-takedown channels built with the majors are working as designed - and that designed is no longer enough.